Publié le : 10/05/2026 à 10:00

Discover the fundamental concepts of Grafana Alloy, the transition from the static Agent to Alloy, its internal architecture based on programmable pipelines, and the different installation methods.

1. What is Grafana Alloy?

Understand the transition from the static Agent to Alloy.

The evolution of the Grafana Agent

Grafana Alloy represents the next generation telemetry collector from Grafana Labs. It succeeds Grafana Agent by adopting a much more flexible approach. Unlike a static agent with a rigid YAML file, Alloy acts as a highly configurable OpenTelemetry distribution, capable of managing metrics, logs, traces, and profiles with great agility.

2. The "Programmable Pipelines" concept

Introduction to declarative syntax.

Dynamic pipelines

Alloy introduces the concept of programmable pipelines using the River configuration language (similar to Terraform's HCL). You no longer just declare targets; you build true data flows where the output of one component becomes the input of another. This allows for complex transformations, conditional routing, and unprecedented scalability.

3. Internal architecture

Components, services, and the Directed Acyclic Graph (DAG).

The DAG at the core of the system

Alloy's internal architecture is based on a Directed Acyclic Graph (DAG). Each configuration block (e.g., discovery, scraping, forwarding) is a component (a node in the graph). Alloy automatically determines the execution order and dependencies between these components. If the configuration changes, only the affected components are reloaded, without restarting the whole process.


Best Practice: Favor splitting your configurations into multiple small River files rather than one huge file. Alloy can load an entire directory, which facilitates organization by team or domain.
Common Mistake: Manually restarting the service (systemctl restart alloy) for every configuration change. Use the hot-reload API instead to apply changes without service interruption.
# Loading an entire directory (more modular)
alloy run /etc/alloy/config.d/

# Hot-reloading the configuration
curl -X POST http://localhost:12345/-/reload

4. Installation

Deployment on Linux, Docker, and Kubernetes.

Deployment options

Alloy is designed to adapt to any environment:

  • Linux (Binary/Systemd): Ideal for classic virtual machines (VMs). Installed via APT/YUM repositories.
  • Docker: An official image is available for quick deployment in simple containerized environments.
  • Kubernetes: Grafana offers an official Helm Chart that allows deploying Alloy as a DaemonSet (for collecting node logs and metrics) or Deployment/StatefulSet (for OTLP ingestion or clustering).
Lien copié dans le presse-papiers !